Bounded Modal Logic

Under the Curry–Howard isomorphism, the syntactic structure of programs can be modeled using birelational Kripke structures equipped with intuitionistic and modal relations. Intuitionistic relations capture scoping through persistence, reflecting the availability of resources from outer scopes, while modal relations model resource isolation introduced for various purposes. Traditional modal languages, however, describe only modal transitions and thus provide limited support for expressing fine-grained control over resource availability. Motivated by this limitation, we introduce \emph{Bounded Modal Logic (\textbf{BML})}, an experimental extension of constructive modal logic whose language explicitly accounts for both intuitionistic and modal transitions. We present a natural-deduction proof system and a Kripke semantics for \textbf{BML}, together with a Curry–Howard interpretation via a corresponding typed lambda-calculus. We establish metatheoretic properties of the calculus, showing that \textbf{BML} forms a well-disciplined logical system. This provides theoretical support for our proposed perspective on fine-grained resource control in programming languages.

💡 Research Summary

The paper introduces Bounded Modal Logic (BML), an extension of constructive modal logic designed to capture both intuitionistic (scope‑inclusion) and modal (resource‑isolation) transitions within a single formal system. Traditional constructive modal logics model only modal transitions, which limits their ability to express fine‑grained resource availability, especially in settings such as multi‑stage programming, distributed computation, or effect systems where code fragments must be valid only under certain scopes.

Core technical contribution
The authors define a birelational Kripke structure—the BML‑structure—as a quintuple ⟨D, ⪯, ⊑, V, !⟩.

• D is a set of “locations” (scopes).
• ⪯ is a preorder representing intuitionistic inclusion (outer scopes are persistently available).
• ⊑ is a second preorder representing modal isolation; it must satisfy the stability condition ⪯ ⊆ ⊑ (equivalently left‑ and right‑stability).
• V maps each atomic proposition to an upward‑closed subset of D.
• ! is a distinguished global scope, the bottom element of ⪯.

On top of this structure the language of BML is built:

A, B ::= p | A → B | □ ⪰ γ A | ∀ γ₁ : ⪰ γ₂ . A

Here γ denotes a classifier, an atomic identifier for a scope. The modality □ ⪰ γ A reads “A can be transported across a modal transition into any scope that has access to the resources of γ via persistence”. The polymorphic classifier quantifier ∀ γ₁ : ⪰ γ₂ . A abstracts over scopes γ₁ that are nested inside a lower bound γ₂.

Proof system
Contexts Γ are sequences of items, each possibly labelled with a classifier. Items include ordinary assumptions A γ, modal delimiters ▶ γ₁ : ⪰ γ₂ (introducing a new modal transition and moving the current position to γ₁), and reverse delimiters ◀ γ (requiring a modal transition in the opposite direction). The current position pos(Γ) is defined inductively and corresponds to a node in the underlying BML‑structure. The natural‑deduction rules extend the usual →‑I/E, □‑I/E, ∀‑I/E with explicit handling of ⪯ and ⊑ via rules such as ⊴‑Refl, ⊴‑Trans, ⊑‑Lift, and the stability‑ensuring rule WF‑◀. This design makes the proof system a direct syntactic counterpart of the Kripke model.

Axioms and derived principles
Figure 2 lists BML analogues of the classic S4 axioms (K, T, 4‑1, 4) together with novel principles:

• Mon and Mon‑1 express the equivalence between □ ⪰ γ A and ∀ γ′ : ⪰ γ . □ ⪰ γ′ A, capturing the persistent nature of classifiers.
• K‑1* shows how a polymorphic classifier quantifier can serve as a converse to K, enabling “inverse” modal reasoning.

Kripke semantics
Satisfaction Γ ⊢ A is defined on BML‑structures. □ ⪰ γ A holds at a world w iff for every world w′ reachable by ⊑ from w, whenever w′ can reach γ via ⪯, the formula A holds at w′. The universal quantifier ranges over all classifiers γ′ that are ≥ γ with respect to ⪯. Soundness and completeness theorems are proved, establishing that the proof system exactly captures the semantics.

Computational interpretation
A corresponding modal λ‑calculus is presented. The term constructors box, unbox, let, and run implement the logical connectives:

• box introduces a value of type □ ⪰ γ A.
• unbox eliminates it, requiring a modal transition.
• let … in … and run model code quotation and execution, mirroring multi‑stage programming.
  The typing rules are in one‑to‑one correspondence with the natural‑deduction rules. The authors prove normalization, strong normalization, and subject reduction, showing that well‑typed BML programs evaluate safely and terminate.

Relation to CS4
In the final section the authors demonstrate that BML strictly generalizes Constructive S4 (CS4). Semantically, a CS4‑structure is a BML‑structure where ⊑ = ⪯. Proof‑theoretically, every CS4 derivation can be simulated in BML, while BML admits additional derivations that exploit classifier quantification and bounded modalities. Thus BML retains the desirable meta‑properties of CS4 while offering richer expressive power for resource‑sensitive typing.

Motivating examples
The paper discusses several scenarios where bounded modalities are essential:

• In MetaML‑style multi‑stage programming, a quoted fragment ⟨sqr 10⟩ should be typed as □ ⪰ γ Int where γ is the scope containing the definition of sqr. The traditional □ Int would incorrectly claim the fragment works under any modal transition.
• Distributed programming may require values that can be sent only to agents satisfying a particular resource classifier.
• Effect systems can restrict effects to specific lexical scopes.

Conclusion
Bounded Modal Logic provides a unified logical framework that simultaneously models persistent scope inclusion and controlled isolation. By integrating classifiers, bounded modalities, and polymorphic quantification over scopes, BML enables fine‑grained, type‑safe reasoning about resource availability in modern programming language designs. The combination of a sound and complete proof system, a Kripke semantics, and a well‑behaved λ‑calculus interpretation makes BML a solid foundation for future research on scoped resource control.